About AI Defense
Defensive AI engineering — guardrails, hardening, response.
What this site covers
AI Defense is about the engineering side of defending LLM systems. It covers the components teams actually assemble under pressure: guardrails, classifiers, filters, and the response paths behind them, together with the failure modes each one has when an attacker is deliberately probing it.
- Guardrail architecture and where to place enforcement
- Prompt-injection and jailbreak detection approaches
- Output filtering and content-safety classifiers
- Audit logging, retention, and what a record has to contain
- Benchmarking guardrails on your own traffic rather than a vendor's
- Refusal behaviour and model hardening
- Mapping controls to the OWASP LLM Top 10 and the EU AI Act
19 articles are published so far. New ones are announced on the RSS feed.
Where to start
Four guides carry most of the weight here. Output filtering architecture for production LLMs is the reference for what leaves the model. LLM audit logging covers what gets recorded and for how long. Monitoring LLM outputs in production covers detecting when any of it stops working, and LLM guardrail benchmarks covers measuring a control on your own traffic instead of trusting a vendor's number.
The site also runs one interactive tool: the Guardrail Gap Analyzer. It takes an application's surface, trust boundary, data sensitivity and hosting model, and returns a layered control matrix with every gap scored by residual risk. It runs entirely in the browser and sends nothing anywhere.
How these articles are produced
Articles here are researched from primary sources: vendor and project documentation, published standards and specifications, research papers and preprints, and measurements published by whoever took them. Drafts are produced with AI assistance and then edited against those cited sources before anything is published.
No article on this site is based on first-hand testing in a private lab, and nothing here should be read as a measurement report of its own. Where a number appears, it comes from a source that is named, so you can check the original instead of taking this site's word for it.
Everything is published under a single editorial byline. That byline is a publishing identity for the site, not a claim about a named individual, and it does not carry professional credentials.
Corrections
Corrections are welcome. If something here is wrong, out of date, or attributed to the wrong source, email hello@aidefense.dev with the page address and what it should say. Substantive corrections are made in the article itself rather than quietly dropped.
How this site is funded
This site currently runs no affiliate links, sponsored posts, display advertising or paid placements. If that changes, the disclosure page will say so.
Contact
Email: hello@aidefense.dev
Site: aidefense.dev
Published by: AI Defense Editorial
See also the privacy policy, the terms of use, and the editorial disclosure.